Agentic Engineering — the Cofoundy operator's knowledge base

teamagentic-engineeringv2agentic-engineeringreference
What this is. The home for how Cofoundy operates when AI is the first reflex, not the last. Everything an operator needs to go from "uses AI but babysits it" → "delegates, orchestrates, and improves the system itself." Onboarding points here.

Agentic engineering is one discipline with layers. Read in this order:

  1. ✓
    Fluency — the map
    The 24 progressive concepts that take you from babysitting to operating. Start here.
  2. 2
    Loop engineering — the how
    Give agents the tools + instructions to verify their own work, run in parallel, and loop in the background. Keyboard out of the hot path.
  3. 3
    Verification · prompt-as-source · multi-agent
    The deeper mechanics — coming as this base grows from real sessions.

The map

Fluency: the AI-Native Curriculum — 24 concepts, operator → system-improver. (Lives under team-onboarding; the fluency layer of agentic engineering.)
Loop engineering: verification loops, multi-agent parallelism, background loops, prompt-as-source. The operating doctrine.
Coming: verification (teach the agent to check its own work) · prompt-as-source control · multi-agent orchestration · the /cto flywheel — distilled from real sessions, append-only.

Start

New operator
Begin with the fluency map, then come back here for loop engineering — how to put agents in loops so they hill-climb to working results on their own.
AI-Native Curriculum →
This base grows from real runs. Every substantive session that surfaces a new agentic-engineering pattern appends here — the same self-improving discipline the docs themselves describe. (This page, like the rest, was authored and published through the docs-ai MCP.)